AI Chatbot Security: How to Evaluate Chatsy
A precise overview of Chatsy's current security boundaries, data flows, available controls, and procurement limitations.
Security claims should be specific enough to verify. This page describes the controls visible in the current Chatsy product and codebase, the data paths a buyer should review, and the assurances Chatsy does not yet provide.
TL;DR:
- Chatsy requires authenticated dashboard sessions and scopes application access to the current user or workspace membership.
- Public embed endpoints validate chatbot access, constrain request bodies, and apply server-side usage and rate-limit checks.
- AI prompts and retrieved context are sent through OpenRouter to the model selected for the agent; review that third-party data path before using sensitive content.
- Custom HTTP and OpenAPI tools can send conversation-derived data to endpoints you configure, so you are responsible for endpoint security, authorization, and data minimization.
- Chatsy does not currently offer SOC 2 Type II reports, HIPAA BAAs, SSO, customer-facing audit-log exports, formal uptime SLAs, dedicated encryption keys, or selectable data-residency regions.
What Chatsy protects in the application
Dashboard APIs require an authenticated session. Workspace and resource checks are applied before users can read or change chatbots, conversations, knowledge bases, mailboxes, live-chat queues, and billing state.
Public chatbot endpoints are intentionally accessible to website visitors, but they still validate the target chatbot and its configuration. Request parsing uses bounded payload helpers on sensitive embed routes, and usage enforcement prevents a workspace from continuing AI responses after its plan credits are exhausted.
These are application controls, not a certification. A production deployment still needs testing for authorization failures, rate limits, abuse, prompt injection, data leakage, and the customer's own website configuration.
Understand the AI data path
Chatsy uses OpenRouter to access supported language and embedding models. A model request can include the system instructions, the visitor's messages, relevant retrieved source passages, tool definitions, and recent conversation context.
Do not place secrets, credentials, payment-card data, protected health information, or other regulated records in a public chatbot. Restrict knowledge sources to content approved for the intended audience. If your policy requires a specific model provider, region, retention commitment, or data-processing term, confirm it in writing before purchase.
Knowledge sources and conversations
Uploaded documents, crawled pages, pasted text, and knowledge-base articles can become retrieval sources for the AI agent. A source is only as safe as its audience and access assumptions. Do not upload private internal material to an agent intended for anonymous public visitors unless every answer derived from that material is safe to disclose.
Chatsy stores conversation history so workspace users can review support activity and continue human handoffs. Use the available deletion controls and contact Chatsy for account-level requests, but do not assume a contractual deletion deadline, cryptographic erasure process, configurable retention period, or legal hold unless it appears in your signed agreement.
Custom tools expand the boundary
Custom HTTP and OpenAPI tools let an AI agent call endpoints supplied by the workspace owner. This is powerful and changes the risk profile:
- Expose the minimum fields and actions needed for the support workflow.
- Authenticate every endpoint independently of the chatbot prompt.
- Validate tool arguments server-side and reject unexpected values.
- Prefer read-only or reversible actions during the first rollout.
- Require human approval for refunds, account changes, eligibility decisions, or other high-impact actions.
- Avoid returning secrets or broad customer records in tool responses.
Chatsy does not make an external endpoint safe merely by calling it through an AI tool. Your team owns the authorization model and business rules of that endpoint.
Current procurement limitations
Do not represent Chatsy as certified or contract-ready for requirements it has not met. At the time of this update, Chatsy does not provide:
- SOC 2 Type II reports or a completed SOC 2 attestation
- HIPAA Business Associate Agreements
- SAML or OIDC single sign-on for the dashboard
- Customer-facing audit-log exports
- FedRAMP, PCI DSS, or industry-specific compliance certification
- On-premises or customer-managed deployment
- Selectable US/EU data-residency regions
- Dedicated customer encryption keys
- A published uptime SLA or multi-region failover commitment
If any item is mandatory, do not deploy regulated or sensitive data while assuming it is on the roadmap. Contact Chatsy, request written evidence, and use a provider that already meets the requirement when necessary.
Buyer verification checklist
Before launch:
- Classify the content and conversation data the chatbot may receive.
- Confirm which users and workspace roles can access each resource.
- Review OpenRouter and the selected model provider as subprocessors.
- Test attempts to access another workspace's resources.
- Test oversized payloads, repeated requests, and malformed tool arguments.
- Review every custom endpoint and remove unnecessary actions.
- Define a human path for sensitive, uncertain, or account-specific requests.
- Verify legal, privacy, retention, residency, accessibility, and contractual requirements with qualified reviewers.
Account practices
Use a unique password, protect the email account used for sign-in, remove former team members promptly, and keep API credentials out of prompts and knowledge sources. Chatsy's dashboard does not currently offer two-factor authentication, so securing the connected email account is particularly important.
Report a suspected vulnerability or data exposure to security@chatsy.app. Do not send customer records, credentials, or exploit payloads containing live sensitive data by ordinary email.
Frequently Asked Questions
Is Chatsy SOC 2 certified?
No. Chatsy does not currently provide a SOC 2 Type II report. Do not state that certification is complete or rely on a projected completion date.
Does Chatsy offer a HIPAA BAA?
No. Do not use Chatsy for protected health information. Healthcare deployments should be limited to reviewed, non-PHI public information and should route sensitive requests to an approved clinical system.
Is two-factor authentication available?
Not currently. Password changes and two-factor authentication are still being prepared. Protect the sign-in email account and limit workspace membership in the meantime.
Can I choose an EU data region?
Not currently as a self-service or contracted product capability. Customers with residency requirements should contact Chatsy before purchase and should not assume a region is available.
Is Chatsy automatically compliant with GDPR or CCPA?
No software configuration by itself makes a customer's deployment compliant. Determine your role, lawful basis, notices, subprocessors, retention, deletion, and data-subject request process with qualified counsel. Request any required contractual documentation before sending personal data.
Is data encrypted?
Chatsy relies on managed hosting, database, and storage services, but this public page does not promise a specific cipher, key-rotation schedule, dedicated key, or coverage for every storage layer. Request written architecture and contractual evidence if a specific encryption control is a procurement requirement.
Contact Chatsy about security requirements